全景视觉球面尺度空间不变特征及匹配方法研究
首发时间:2015-12-07
摘要:折光全景视觉成像系统是近年来机器人视觉领域的研究热点之一,传统针孔透视投影模型不符合该类型相机光学成像性质。本文研究了单视点球面投影机理解决全景相机成像建模问题,在此基础上研究球面尺度空间构建方法,球面尺度局部描述子,以及球面尺度不变特征提取算法,在跨尺度空间进行全景图像特征点的提取及特征匹配。实验表明,球面尺度不变特征提取算法能解决全景相机图像高度畸变情况下的图像特征匹配问题。
For information in English, please click here
Spherical Scale Space Invariant Feature and Matching Method of Omnidirectional camera
Abstract:Omni-directional vision system is one of the research hotspots in the field of robot vision in recent years, the traditional perspective projection method doesn't accord with its optical imaging properties. Therefore, in this paper, a single viewpoint projection mechanism is used to construct camera model, and spherical scale space invariant feature, local spherical scale descriptor, and feature extraction algorithm in multiscale space are investigate accordingly. The experimental results show that the algorithm can solve the problem of image feature matching given the high distortion of panoramic image.
Keywords: robot vision spherical scale space invarance omni-directional vision feature matching
论文图表:
引用
No.4667005103831814****
同行评议
共计0人参与
勘误表
全景视觉球面尺度空间不变特征及匹配方法研究
评论
全部评论